On 3/29/07, Tony Hansen <tony(_at_)att(_dot_)com> wrote:
<implementor's hat>
Side note: An implementation is free to respond with 4xx (e.g., 421,
451 or 452) to the MAIL FROM whenever it wants to shut down the
connection. And some implementations *do* choose to do so because of
local policy decisions.
</implementor's hat>
Recently I've encountered servers that respond with 221 rather than
250 at the end of the DATA phase (after <CRLF>.<CRLF>) and then
immediately close the connection. (This may not be a
recently-developed practice, but I noticed it for the first time last
week in an SMTP client log when tracking down a delivery problem.)
It would appear this is meant as an indication that the server intends
to accept exactly one message and doesn't care to wait for a QUIT.
I'm not sure what, if anything, could be put into the document about
this. Perhaps guidance as to how clients should react when the server
wanders off-protocol? In the case I was investigating the client
treated the situation as a 4yz and retried the message later, leading
to a lot of copies being delivered and (fortunately just one) irate
recipient.
(I've also seen 221-and-close responses to EHLO, which seems entirely bizarre.)